FireMV 2200 vs CustomEDID

Alex Deucher alexdeucher at gmail.com
Sun Oct 4 21:34:41 PDT 2009


2009/10/4 Csillag Kristof <csillag.kristof at united-consult.hu>:
> Dear All,
>
> It turned out that the new CustomEDID feature does not always work as
> expected. After some trial-and-error, I have done some systematic testing.
>
> I have tested my integrated Radeon X1250 chipset and a PCI FireMV 2200
> card with version 6.12.3 (with my CustomEDID patch) and the current GIT
> master (with Dave's CustomEDID patch), with a CRT monitor (well,
> actually with the analog port of an LCD), with a proper LCD monitor and
> with a flaky (DDC-broken) LCD monitor, with all possible combinations of
> driver version, card, socket and monitor.
>
> (I have access to a FireMV 2400 card too, but currently I don't have the
> time to run all these tests on in.)
>
> Here are the test results:
>
> ----------------------------------------------------------------------
>
> 1. Radeon X1250
>
> 1.1.1. Proper LCD on Socket 1 - Port1 - DVI-0 - INTERNAL_LVTM1 - OK
> 1.1.2. Flaky LCD on Socket 1 - Port1 - DVI-0 - INTERNAL_LVTM1 - OK
> 1.2.1. CRT on Socket 2 - Port0 - VGA - INTERNAL_KLDSCP_DAC1    - OK
>
> 2. FireMV 2200
>
> 2.1.1. CRT on Socket 1 - Port0 - DVI-1 - INTERNAL_DAC1    - OK
> 2.1.2. Proper LCD on Socket 1 - Port0 - DVI-1 - INTERNAL_DVO1 - OK
> 2.1.3.1. Flaky LCD on Socket 1 - Port0 - DVI-1 - INTERNAL_DVO1 - No picture
> 2.1.3.2. Flaky LCD on Socket 1 - Port0 - DVI-1 - INTERNAL_DVO1, Proper
> LCD connected at X start - OK
> 2.2.1. CRT on Socket 2 - Port1 - DVI-0 - INTERNAL_DAC2    - No picture
> 2.2.2. Proper LCD on Socket 2 - Port1 - DVI-0 - INTERNAL_TMDS1 - OK
> 2.2.3.1. Flaky LCD on Socket 2 - Port1 - DVI-0 - INTERNAL_TMDS1 - No picture
> 2.2.3.2. Flaky LCD on Socket 2 - Port1 - DVI-0 - INTERNAL_TMDS1, Proper
> LCD connected at X start - No picture
>
> Notes:
>  - The the versions of the driver behaved identical in all test cases,
>    so I have not repeated the results.
>  - When working with the Flaky LCD, the CustomEDID option was always
>    used, except I state otherwise.)
>
>
> Findings:
>
>  - The FireMV card fails output analog data to the CRT on one of the
> sockets.
>   According to X log, everything is set up properly, but still,
>   there is no image on the monitor.
>   (See Xorg.log.CRT_failure)
>
>  - On the FireMV 2200 PCI card, the CustomEDID options fails the enforce
>   the requested video mode: according to the X log, everything is set up
>   properly, but still, there is no image on the monitor.
>   The same feature works properly with the Radeon X1250 chipset.
>   Strangely, I can make CustomEDID work on one of the heads by connecting
>   a proper LCD monitor when starting X, and then switching back to the
>   flaky one. (This works with one head, but not with the other.)
>
>   So, it seems that despite using CustomEDID, this card is still doing some
>   low-level monitor detection, and disables the output when it senses no
>   connected device.
>
>   (See Xorg.log.CustomEDID_failure)
>
>
> Do you have any idea how to solve any of those problems?

One problem is that the custom EDID you are using is grabbed from the
analog vga cable from your monitor.  As such it has the analog input
set in the EDID:
(II) RADEON(0): EDID data from the display on output: DVI-1
----------------------
(II) RADEON(0): Manufacturer: SAM  Model: 1cf  Serial#: 1112683057
(II) RADEON(0): Year: 2006  Week: 41
(II) RADEON(0): EDID Version: 1.3
(II) RADEON(0): Analog Display Input,  Input Voltage Level: 0.700/0.300 V
                         ^^^^^^^^^^^^

The driver uses this bit to decide whether to drive the monitor on A
DVI-I port with the analog or digital transmitter.  To work properly
using the digital transmitter you need to edit the EDID so it
specifies a digital input.  The fact that it works in some cases is
probably due to luck and the driver not turning off the digital in
some cases.

Alex



More information about the xorg mailing list